WHY I AM
OPTIMISTIC
Patrick H. Winston
MIT Artificial Intelligence
Laboratory
The salients

Applications side:
We have already won

Science side:
We are bound to win
The Applications
We have expanded our frontiers
Lots of good
Lots of people
A little good for a lot of people
A little good for a lot of people
A lot of good for a few people
A lot of good for a lot of people
We have exemplars of all kinds

Large software companies

Large entertainment companies

Companies with huge IPOs

Multidimensional multinationals

A multitude of small companies
We were a one-horse field
Rule chaining
Inheritance
Now we ride many horses
Neural nets
Rule chaining
Inheritance
Generate and test
Constraint propagation
Search
Tree building
Genetic algorithms
Bayes nets
Agents
Learning
And not just reasoning horses



Vision
Language and speech
Infrastructure
We had a Pyrrhic victory
Network
Memory
Tapes
IO
Cables
Power
Disk
We learned negative lessons
Nobody cares about saving money


Using cutting edge technology
To replace expensive experts
We learned positive lessons
Everybody cares about



New revenues
Saving a mountain of money
Increasing competitiveness
We changed the business model
Replaces
Expensive
People
Ferrets
Blunder stoppers
Novices
Experts
Saves
Mountains
Of Money
Creates
New
Revenue
The critic and the billionaire
What’s next: connections
People
Human
Computer
Interaction
Computers
Information
Access
Enhanced
Reality
Intelligent Structures
Useful robots
Physical World
Global Net
The click-in phenomenon


The fax machine
The world wide web
The Science
Shrobe’s point


Applications drive science
Unless they all look alike
Atkeson’s point


We could move to the center
But, we might be kidding ourselves
My point



AI is applied computer science
Much energy wonderfully used
But consequently diverted
A 100 year enterprise
Molecular Biology
1900
1950
2000
2050
Artificial Intelligence
Why we are the way we are
Powerful Ideas
Models of Thinking
Reflection…Biology…Psychology
The standard paradigm
The Intelligent
Reasoner
Input/Output Channels
Vision
Language
The dawn age
x
(1 - x
4
2
)
5 /2
dx
What went wrong?




We think with our eyes
We think with our mouths
We think with our hands
Each faculty helps the others
What is the evidence?

Armchair psychology

Clues from the brain
Armchair psychology

Hillis’s observation on the value
of talking to yourself

Everyone’s observation on the
value of drawing a sketch.
From brain scanning
Intelligence is in the I/O
The Explanation
Linguistic
Reasoner
Motor
Reasoner
Visual
Reasoner
Is it time to start over?



An I/O oriented paradigm
Essentially free computation
Important, inspiring allies
From brain rewiring
From watching infants
Is it time to start over?




An I/O oriented paradigm
Essentially free computation
Important, inspiring allies
Accumulation of powerful ideas
Six powerful ideas






Recreated condition
One-shot learning
Memory is cheap
Change matters
Survival of the smallest
Bi-directional search
Recreated condition: Minsky
*---------------------------------P
|
/
| K-line
\
P
|
P
/
*-------------->
\
P
|
P
/
*--------->
\
P
P
/
\
P
/
P
/
P
\
/
P
P
/
P
\
P
\
/
P
P
/
P
P
P
/
P
\
\
\
P
\
P
/
P
/
P
\
/
P
P
\
P
\
/
P
P
P
\
/
P
P
\
P
One-shot: Yip and Sussman
Rule
Memory
ae
Time
p
Word
Memory
l
One-shot: Yip and Sussman
Rule
Memory
ae
p
Time
l
Word
Memory
z
One-shot: Yip and Sussman
100%
Accuracy
Trials
500
Memory is cheap: Atkeson
Atkeson’s practice tables
Atkeson’s practice results
Feedback only
One stored trajectory
Three stored trajectories
Change matters: Borchardt
A
Appear
A
Appear
D
Disappear
D
Disappear
Change
Change
Decrease
Decrease
Increase
Increase
Borchardt’s ladder diagrams
Distance
D
A
Speed
D
A
A
A
A
Contact
T1
T2
T3
T4
Survival of the smallest: Kirby
Kirby’s phase transitions
Coverage
Time
Bi-directional search: Ullman
Model
Image
Joyous inferences



Powerful ideas
Marvelous engineering
Essential alliances
What about …



Bayes and Markov
Neural nets and connectionism
Logic
WHAT WE
MUST NOT DO
Loose our faith



It will take 300 years
All the low hanging fruit is gone
We shouldn’t make predictions
Waste time arguing



Is it possible?
Is it successful?
Is it really AI?
Squander our capital




One thousand people
10% interested in the science side
10% actually working on it
10% of the time
WHAT WE
SHOULD DO
Human Intelligence Enterprise





Vision, language, motor
Free hardware
Clues from the brain
Powerful ideas
Conceive and test models
The Human Intelligence Enterprise
Linguistic
Reasoner
Visual
Reasoner
Motor
Reasoner
Why we should do it


It can only be done once
Revolutionary applications
The Human Intelligence Enterprise
Linguistic
Reasoner
Visual
Reasoner
Motor
Reasoner
What we should ask






The Human Intelligence Enterprise
Linguistic
Reasoner
Why do we have discrete words?
What do our inner agents say?
How do they learn what to say?
Do we see what chimps see?
How did our faculties evolve?
Why can’t we all play the piano?
Visual
Reasoner
Motor
Reasoner
So, here is why I’m optimistic



The Human Intelligence Enterprise
Linguistic
Reasoner
Visual
Reasoner
Nothing could possibly be more
fun, exciting, rewarding, and
glorious, than …
Applications that really matter
Figuring out our own intelligence
Motor
Reasoner
Descargar

AAAI 99 Keynote Address